# Healthcare Data Analysis and Disease Prediction ## Project Overview In this project, we will analyze a healthcare dataset to identify patterns and predict the likelihood of a particular disease based on various features such as age, blood pressure, and lifestyle factors. We will implement classification algorithms to make predictions and use various metrics to evaluate our models.

# Import necessary libraries  
import pandas as pd  
import numpy as np  
import seaborn as sns  
import matplotlib.pyplot as plt  
from sklearn.model_selection import train_test_split, cross_val_score  
from sklearn.preprocessing import StandardScaler  
from sklearn.metrics import confusion_matrix, classification_report, roc_auc_score, roc_curve  
from sklearn.ensemble import RandomForestClassifier  
from sklearn.svm import SVC  
from sklearn.tree import DecisionTreeClassifier  
from xgboost import XGBClassifier 
import xgboost as xgb
from sklearn.feature_selection import RFE  
from sklearn.linear_model import LassoCV  

## Load the Dataset We will load the healthcare dataset for analysis. Make sure to download the dataset and update the path below.

# Load the dataset (update the path to your dataset)  
data = pd.read_csv("C:\\Users\\dongr\\OneDrive\\Desktop\\new proj\\diabetes.csv")

# Display the first few rows of the dataset  
data.head()

	id	chol	stab.glu	hdl	ratio	glyhb	location	age	gender	height	weight	frame	bp.1s	bp.1d	bp.2s	bp.2d	waist	hip	time.ppn
0	1000	203.0	82	56.0	3.6	4.31	Buckingham	46	female	62.0	121.0	medium	118.0	59.0	NaN	NaN	29.0	38.0	720.0
1	1001	165.0	97	24.0	6.9	4.44	Buckingham	29	female	64.0	218.0	large	112.0	68.0	NaN	NaN	46.0	48.0	360.0
2	1002	228.0	92	37.0	6.2	4.64	Buckingham	58	female	61.0	256.0	large	190.0	92.0	185.0	92.0	49.0	57.0	180.0
3	1003	78.0	93	12.0	6.5	4.63	Buckingham	67	male	67.0	119.0	large	110.0	50.0	NaN	NaN	33.0	38.0	480.0
4	1005	249.0	90	28.0	8.9	7.72	Buckingham	64	male	68.0	183.0	medium	138.0	80.0	NaN	NaN	44.0	41.0	300.0

# Get basic information about the dataset  
print(data.shape)  # Number of rows and columns  
print(data.info())  # Data types and non-null counts  
print(data.describe())  # Summary statistics  

# Check for missing values  
missing_values = data.isnull().sum()  
print("Missing Values:\n", missing_values[missing_values > 0])

(403, 19)
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 403 entries, 0 to 402
Data columns (total 19 columns):
 #   Column    Non-Null Count  Dtype  
---  ------    --------------  -----  
 0   id        403 non-null    int64  
 1   chol      402 non-null    float64
 2   stab.glu  403 non-null    int64  
 3   hdl       402 non-null    float64
 4   ratio     402 non-null    float64
 5   glyhb     390 non-null    float64
 6   location  403 non-null    object 
 7   age       403 non-null    int64  
 8   gender    403 non-null    object 
 9   height    398 non-null    float64
 10  weight    402 non-null    float64
 11  frame     391 non-null    object 
 12  bp.1s     398 non-null    float64
 13  bp.1d     398 non-null    float64
 14  bp.2s     141 non-null    float64
 15  bp.2d     141 non-null    float64
 16  waist     401 non-null    float64
 17  hip       401 non-null    float64
 18  time.ppn  400 non-null    float64
dtypes: float64(13), int64(3), object(3)
memory usage: 59.9+ KB
None
                 id        chol    stab.glu         hdl       ratio  \
count    403.000000  402.000000  403.000000  402.000000  402.000000   
mean   15978.310174  207.845771  106.672457   50.445274    4.521642   
std    11881.122124   44.445557   53.076655   17.262626    1.727886   
min     1000.000000   78.000000   48.000000   12.000000    1.500000   
25%     4792.500000  179.000000   81.000000   38.000000    3.200000   
50%    15766.000000  204.000000   89.000000   46.000000    4.200000   
75%    20336.000000  230.000000  106.000000   59.000000    5.400000   
max    41756.000000  443.000000  385.000000  120.000000   19.299999   

            glyhb         age      height      weight       bp.1s       bp.1d  \
count  390.000000  403.000000  398.000000  402.000000  398.000000  398.000000   
mean     5.589769   46.851117   66.020101  177.592040  136.904523   83.321608   
std      2.242595   16.312333    3.918515   40.340666   22.741033   13.589227   
min      2.680000   19.000000   52.000000   99.000000   90.000000   48.000000   
25%      4.380000   34.000000   63.000000  151.000000  121.250000   75.000000   
50%      4.840000   45.000000   66.000000  172.500000  136.000000   82.000000   
75%      5.600000   60.000000   69.000000  200.000000  146.750000   90.000000   
max     16.110001   92.000000   76.000000  325.000000  250.000000  124.000000   

            bp.2s       bp.2d       waist         hip     time.ppn  
count  141.000000  141.000000  401.000000  401.000000   400.000000  
mean   152.382979   92.524823   37.900249   43.039900   341.250000  
std     21.712952   11.555198    5.729313    5.656713   309.540953  
min    110.000000   60.000000   26.000000   30.000000     5.000000  
25%    138.000000   84.000000   33.000000   39.000000    90.000000  
50%    149.000000   92.000000   37.000000   42.000000   240.000000  
75%    161.000000  100.000000   41.000000   46.000000   517.500000  
max    238.000000  124.000000   56.000000   64.000000  1560.000000  
Missing Values:
 chol          1
hdl           1
ratio         1
glyhb        13
height        5
weight        1
frame        12
bp.1s         5
bp.1d         5
bp.2s       262
bp.2d       262
waist         2
hip           2
time.ppn      3
dtype: int64

non_numeric_columns = data.select_dtypes(exclude=[np.number]).columns  
print(non_numeric_columns)

Index(['location', 'gender', 'frame'], dtype='object')

data[non_numeric_columns] = data[non_numeric_columns].apply(pd.to_numeric, errors='coerce')

data.drop(columns=non_numeric_columns, inplace=True)

# Fill missing values or drop rows/columns (example filling with mean)  
data.fillna(data.mean(), inplace=True)

# Example: if you have Weight and Height, calculate BMI (adjust according to your dataset)  
if 'Weight' in data.columns and 'Height' in data.columns:  
    data['BMI'] = data['Weight'] / (data['Height'] ** 2)

# Create a list of numerical features to scale  
numerical_features = ['height', 'weight']    

scaler = StandardScaler()  
data[numerical_features] = scaler.fit_transform(data[numerical_features])

# Example: Histogram of a feature  
sns.histplot(data['age'], bins=30)  
plt.title('Distribution of age')  
plt.xlabel('age')  
plt.ylabel('Frequency')  
plt.show()

plt.figure(figsize=(10, 8))  
sns.heatmap(data.corr(), annot=True, cmap='coolwarm')  
plt.title('Feature Correlation Heatmap')  
plt.show()

# visualizing through abscatterplot diagram 
sns.scatterplot(data=data, x='chol', y='hdl', hue='age')  
plt.title('Cholesterol vs HDL with Age as Hue')  
plt.xlabel('Cholesterol')  
plt.ylabel('HDL')  
plt.legend(title='Age')  
plt.show()

#pair plot visualization
sns.pairplot(data, hue='age')  
plt.show()

# Define features and target variable  
features = data.columns.tolist()  # Adjust based on your dataset and selected features  
target = 'age'    

X = data[features]  
y = data[target]  

# Apply RFE  
model = RandomForestClassifier()  
selector = RFE(model, n_features_to_select=5)  # Adjust number as needed  
selector = selector.fit(X, y)  

# Store selected features  
selected_features_rfe = [features[i] for i in range(len(features)) if selector.support_[i]]  
print("Selected Features using RFE: ", selected_features_rfe)

Selected Features using RFE:  ['id', 'chol', 'hdl', 'glyhb', 'age']

# Define models
models = {
    'SVM': SVC(probability=True, random_state=42),
    'Decision Tree': DecisionTreeClassifier(random_state=42),
    'Random Forest': RandomForestClassifier(random_state=42),
    'xgboost': xgb.XGBClassifier(use_label_encoder=False, eval_metric='logloss', random_state=42)
}

# Feature Scaling 
scaler = StandardScaler()
data_scaled = scaler.fit_transform(data.drop('age',axis=1))
data_scaled = pd.DataFrame(data_scaled, columns=data.columns[:-1])

X = data_scaled
y = data['age']
from sklearn.feature_selection import RFE
from sklearn.linear_model import LogisticRegression

# Initialize the Logistic Regression model
model = LogisticRegression(max_iter=1000)  # Ensure convergence

# Initialize RFE with the model and specify the number of features to select
rfe = RFE(estimator=model, n_features_to_select=5)  # Use keyword argument for n_features_to_select

# Fit RFE
fit = rfe.fit(X, y)

# Output selected features
print("Selected Features: ", fit.support_)
print("Feature Ranking: ", fit.ranking_)

Selected Features:  [False False False  True  True  True False  True  True False False False
 False False False]
Feature Ranking:  [ 4 10  7  1  1  1  9  1  1  5  6  3  2  8 11]

X_train, X_test, y_train, y_test = train_test_split(data[selected_features_rfe], data[target], test_size=0.2, random_state=42)

# Create a dictionary of models  
models = {  
    "SVM": SVC(probability=True),  
    "Decision Tree": DecisionTreeClassifier(),  
    "Random Forest": RandomForestClassifier()
}
# Fit each model to the training data and print a success message  
for name, model in models.items():  
    model.fit(X_train, y_train)  
    print(f"{name} trained successfully.")

SVM trained successfully.
Decision Tree trained successfully.
Random Forest trained successfully.

for name, model in models.items():  
    scores = cross_val_score(model, X_train, y_train, cv=5, scoring='accuracy')  
    print(f"{name} Accuracy: {scores.mean():.2f} +/- {scores.std():.2f}")

C:\Users\dongr\anaconda3\Lib\site-packages\sklearn\model_selection\_split.py:737: UserWarning: The least populated class in y has only 1 members, which is less than n_splits=5.
  warnings.warn(

SVM Accuracy: 0.03 +/- 0.01
Decision Tree Accuracy: 0.91 +/- 0.03

C:\Users\dongr\anaconda3\Lib\site-packages\sklearn\model_selection\_split.py:737: UserWarning: The least populated class in y has only 1 members, which is less than n_splits=5.
  warnings.warn(
C:\Users\dongr\anaconda3\Lib\site-packages\sklearn\model_selection\_split.py:737: UserWarning: The least populated class in y has only 1 members, which is less than n_splits=5.
  warnings.warn(

Random Forest Accuracy: 0.53 +/- 0.03

## Conclusion and Recommendations Based on the analysis and model performance, we recommend [decision tree with more accuracy]andThe Random Forest model performed best with an AUC of [0.53]. Potential improvements include [has to train with xg boost].